[lolinder]

The big problem with the first design was that people don't have a set of "favorite" trails that are all weighted equally. People have a favorite trail, and a backup trail, and a backup trail for that one, and so on. It's a list, not a set.

The first design assumes that if I open the app I must be interested in a random trail from my unsorted favorites set that happens to have good conditions. The second design does better not just because it helps find close trails—that's just one factor in my preference—it does better because it helps me to find reports on a specific trail. If those aren't favorable I can go to my backup trail and then on down my list.

As others have said, this isn't a failure to match the user's mental models or to provide enough of a "puzzle". It's a failure to understand the user's goal, the job they need your app to do. I would bet that a design that allowed users to pin trails to a dashboard (similar to a home screen) would perform similarly well or better than the map.